ChatGPT for Mac
Run the ChatGPT Mac app on your OpenLLM models with the OpenLLM ChatGPT launcher that openllm chatgpt installs, with or without your ChatGPT sign-in.
| Command | openllm chatgpt |
| Mode | Persistent launcher app. Your ~/.codex is never modified. |
| Talks to | OpenAI Responses API at <gateway>/v1, through the app's built-in Codex engine |
| Default model | ultra |
| Platform | macOS only |
| Remove | openllm chatgpt uninstall |
Before you begin
- Install the OpenLLM CLI on your Mac. See Install.
- Install the ChatGPT app at
/Applications/ChatGPT.appfromhttps://chatgpt.com/download. - Have at least one model available. See Providers.
Set up with openllm chatgpt
openllm chatgptThe command asks how to handle your ChatGPT account:
| Choice | What happens |
|---|---|
| With ChatGPT sign-in (default) | Models run through OpenLLM. Dictation, ChatGPT connectors and apps, and Work mode keep working on your ChatGPT account and plan. |
| Without ChatGPT sign-in | The app runs with no ChatGPT account. Those features are turned off. |
To skip the question, pass --with-chatgpt-auth or --without-chatgpt-auth. Your choice is remembered; run the command again with the other flag to change it.
The command installs /Applications/OpenLLM ChatGPT.app and records it in ~/.openllm/clients/chatgpt.json. It does not start ChatGPT.
What happens when you open OpenLLM ChatGPT
Open OpenLLM ChatGPT from Spotlight or Launchpad. Each time, the launcher:
- Picks a gateway: your local daemon at
http://127.0.0.1:8787if it answers, otherwisehttps://www.openllm.sh. - Fetches your model catalog to
~/.openllm/clients/chatgpt/models.json. - Writes
~/.openllm/clients/chatgpt/codex-shim, which runs the app's built-in Codex engine with the same overrides asopenllm codex, theopenllmMCP server, and memory hooks. - Quits ChatGPT if it is running, then reopens it with
CODEX_CLI_PATHpointing at the shim and your key inOPENLLM_API_KEY.
Your chats and ~/.codex stay as they are. To use ChatGPT without OpenLLM, quit it and open ChatGPT as usual.
Choose models
The default model is ultra. The app's model picker lists ultra, plus, lite, and every model you have activated. To pin a model, pick it there. To change what the aliases resolve to, edit the chains in Configuration. See Routing and fallback.
The memory hooks recall before each prompt and save after each turn. If ChatGPT asks you to approve them, approve once. See Memory.
Who pays for a request
Model requests use your OpenLLM key, and the model that serves them decides who pays. A ChatGPT subscription hop runs on your own ChatGPT login through the daemon on one of your devices. An API-key hop bills your provider account. Included openllm/ models count against your plan's allowance. With ChatGPT sign-in kept, dictation and connectors use your ChatGPT plan, not OpenLLM. See Providers.
Check that it works
openllm chatgpt statusThis prints one JSON line. "installed": true means the launcher is in place; chatgpt_auth shows your sign-in choice. Open OpenLLM ChatGPT, send a message, and check Overview in the app.
Common problems
| Symptom | Cause | Fix |
|---|---|---|
ChatGPT.app not found at /Applications/ChatGPT.app. | The ChatGPT app is not installed there | Install it, then run openllm chatgpt again |
| Alert: Unsupported ChatGPT version | A ChatGPT update moved its built-in engine | Run openllm update, then open the launcher again |
| Alert: OpenLLM needs an API key | No key is saved | Run openllm chatgpt in a terminal and paste your key |
already exists and isn't an OpenLLM launcher — refusing to replace it. | Another app has the same name | Move or rename that app, then run the command again |
| The microphone button is missing | You chose Without ChatGPT sign-in | Run openllm chatgpt --with-chatgpt-auth, then reopen the launcher |
| The picker shows only OpenAI's models | ChatGPT was opened directly | Quit ChatGPT and open OpenLLM ChatGPT |
For anything else, see Troubleshooting.
Remove it
openllm chatgpt uninstallThis removes the launcher app, ~/.openllm/clients/chatgpt/, and the record. If ChatGPT is running through OpenLLM, quit it and open it normally.
Last updated on
Raycast
Add your OpenLLM models and the ultra, plus and lite aliases to Raycast AI on macOS with openllm raycast, and remove them with openllm raycast uninstall.
DeepSeek Harness
Route DeepSeek Harness (dsh) through the OpenLLM daemon on your machine by adding the OpenLLM bundle to a dsh profile with the dsh plugin command.